2010-07-15 15:13:44

by Catalin Marinas

[permalink] [raw]
Subject: [PATCH] kmemleak: Add support for NO_BOOTMEM configurations

With commits 08677214 and 59be5a8e, alloc_bootmem()/free_bootmem() and
friends use the early_res functions for memory management when
NO_BOOTMEM is enabled. This patch adds the kmemleak calls in the
corresponding code paths for bootmem allocations.

Signed-off-by: Catalin Marinas <[email protected]>
Cc: Pekka Enberg <[email protected]>
Cc: Yinghai Lu <[email protected]>
Cc: H. Peter Anvin <[email protected]>
---
kernel/early_res.c | 6 ++++++
mm/page_alloc.c | 5 +++++
2 files changed, 11 insertions(+), 0 deletions(-)

diff --git a/kernel/early_res.c b/kernel/early_res.c
index 31aa933..7bfae88 100644
--- a/kernel/early_res.c
+++ b/kernel/early_res.c
@@ -7,6 +7,8 @@
#include <linux/bootmem.h>
#include <linux/mm.h>
#include <linux/early_res.h>
+#include <linux/slab.h>
+#include <linux/kmemleak.h>

/*
* Early reserved memory areas.
@@ -319,6 +321,8 @@ void __init free_early(u64 start, u64 end)
struct early_res *r;
int i;

+ kmemleak_free_part(__va(start), end - start);
+
i = find_overlapped_early(start, end);
r = &early_res[i];
if (i >= max_early_res || r->end != end || r->start != start)
@@ -333,6 +337,8 @@ void __init free_early_partial(u64 start, u64 end)
struct early_res *r;
int i;

+ kmemleak_free_part(__va(start), end - start);
+
if (start == end)
return;

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 431214b..68319dd 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3659,6 +3659,11 @@ void * __init __alloc_memory_core_early(int nid, u64 size, u64 align,
ptr = phys_to_virt(addr);
memset(ptr, 0, size);
reserve_early_without_check(addr, addr + size, "BOOTMEM");
+ /*
+ * The min_count is set to 0 so that bootmem allocated blocks
+ * are never reported as leaks.
+ */
+ kmemleak_alloc(ptr, size, 0, 0);
return ptr;
}


2010-07-15 17:21:52

by Yinghai Lu

[permalink] [raw]
Subject: Re: [PATCH] kmemleak: Add support for NO_BOOTMEM configurations

should be for 2.6.34 and 2.6.35.

Thanks

Yinghai

On 07/15/2010 07:58 AM, Catalin Marinas wrote:
> With commits 08677214 and 59be5a8e, alloc_bootmem()/free_bootmem() and
> friends use the early_res functions for memory management when
> NO_BOOTMEM is enabled. This patch adds the kmemleak calls in the
> corresponding code paths for bootmem allocations.
>
> Signed-off-by: Catalin Marinas <[email protected]>
> Cc: Pekka Enberg <[email protected]>
> Cc: Yinghai Lu <[email protected]>
> Cc: H. Peter Anvin <[email protected]>
> ---
> kernel/early_res.c | 6 ++++++
> mm/page_alloc.c | 5 +++++
> 2 files changed, 11 insertions(+), 0 deletions(-)
>
> diff --git a/kernel/early_res.c b/kernel/early_res.c
> index 31aa933..7bfae88 100644
> --- a/kernel/early_res.c
> +++ b/kernel/early_res.c
> @@ -7,6 +7,8 @@
> #include <linux/bootmem.h>
> #include <linux/mm.h>
> #include <linux/early_res.h>
> +#include <linux/slab.h>
> +#include <linux/kmemleak.h>
>
> /*
> * Early reserved memory areas.
> @@ -319,6 +321,8 @@ void __init free_early(u64 start, u64 end)
> struct early_res *r;
> int i;
>
> + kmemleak_free_part(__va(start), end - start);
> +
> i = find_overlapped_early(start, end);
> r = &early_res[i];
> if (i >= max_early_res || r->end != end || r->start != start)
> @@ -333,6 +337,8 @@ void __init free_early_partial(u64 start, u64 end)
> struct early_res *r;
> int i;
>
> + kmemleak_free_part(__va(start), end - start);
> +
> if (start == end)
> return;
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 431214b..68319dd 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -3659,6 +3659,11 @@ void * __init __alloc_memory_core_early(int nid, u64 size, u64 align,
> ptr = phys_to_virt(addr);
> memset(ptr, 0, size);
> reserve_early_without_check(addr, addr + size, "BOOTMEM");
> + /*
> + * The min_count is set to 0 so that bootmem allocated blocks
> + * are never reported as leaks.
> + */
> + kmemleak_alloc(ptr, size, 0, 0);
> return ptr;
> }
>

2010-07-16 08:49:58

by Catalin Marinas

[permalink] [raw]
Subject: Re: [PATCH] kmemleak: Add support for NO_BOOTMEM configurations

On Thu, 2010-07-15 at 18:16 +0100, Yinghai Lu wrote:
On 07/15/2010 07:58 AM, Catalin Marinas wrote:
> > With commits 08677214 and 59be5a8e, alloc_bootmem()/free_bootmem() and
> > friends use the early_res functions for memory management when
> > NO_BOOTMEM is enabled. This patch adds the kmemleak calls in the
> > corresponding code paths for bootmem allocations.
> >
> > Signed-off-by: Catalin Marinas <[email protected]>
> > Cc: Pekka Enberg <[email protected]>
> > Cc: Yinghai Lu <[email protected]>
> > Cc: H. Peter Anvin <[email protected]>
>
> should be for 2.6.34 and 2.6.35.

I only noticed it recently since we don't have NO_BOOTMEM on ARM.
Anyway, I'll cc [email protected] when pushing the patch to Linus.

Should I consider your reply as an acked-by?

Thanks.

--
Catalin

2010-07-16 09:01:50

by Pekka Enberg

[permalink] [raw]
Subject: Re: [PATCH] kmemleak: Add support for NO_BOOTMEM configurations

Catalin Marinas wrote:
> With commits 08677214 and 59be5a8e, alloc_bootmem()/free_bootmem() and
> friends use the early_res functions for memory management when
> NO_BOOTMEM is enabled. This patch adds the kmemleak calls in the
> corresponding code paths for bootmem allocations.
>
> Signed-off-by: Catalin Marinas <[email protected]>
> Cc: Pekka Enberg <[email protected]>
> Cc: Yinghai Lu <[email protected]>
> Cc: H. Peter Anvin <[email protected]>

Acked-by: Pekka Enberg <[email protected]>

2010-07-16 18:48:17

by Yinghai Lu

[permalink] [raw]
Subject: Re: [PATCH] kmemleak: Add support for NO_BOOTMEM configurations

On 07/16/2010 01:46 AM, Catalin Marinas wrote:
> On Thu, 2010-07-15 at 18:16 +0100, Yinghai Lu wrote:
> On 07/15/2010 07:58 AM, Catalin Marinas wrote:
>>> With commits 08677214 and 59be5a8e, alloc_bootmem()/free_bootmem() and
>>> friends use the early_res functions for memory management when
>>> NO_BOOTMEM is enabled. This patch adds the kmemleak calls in the
>>> corresponding code paths for bootmem allocations.
>>>
>>> Signed-off-by: Catalin Marinas <[email protected]>
>>> Cc: Pekka Enberg <[email protected]>
>>> Cc: Yinghai Lu <[email protected]>
>>> Cc: H. Peter Anvin <[email protected]>
>>
>> should be for 2.6.34 and 2.6.35.
>
> I only noticed it recently since we don't have NO_BOOTMEM on ARM.
> Anyway, I'll cc [email protected] when pushing the patch to Linus.
>
> Should I consider your reply as an acked-by?
>

Yes.

Thanks

Yinghai